3GPP report skeleton

g3rd Generation Partnership Project;

Technical Specification Group Radio Access Network;

Dual-Cell HSDPA operation;

(Release 8)

The present document has been developed within the 3rd Generation Partnership Project (3GPP TM) and may be further elaborated for the purposes of 3GPP.

The present document has not been subject to any approval process by the 3GPP Organizational Partners and shall not be implemented.

This Specification is provided for future development work within 3GPP only. The Organizational Partners accept no liability for any use of this Specification.Specifications and reports for implementation of the 3GPP TM system should be obtained via the 3GPP Organizational Partners' Publications Offices.

3GPP TR 25.825 V1.0.0 (2008-05)Technical Report

Keywords

UMTS, radio, packet mode

3GPP

Postal address

3GPP support office address

650 Route des Lucioles - Sophia Antipolis

Valbonne - FRANCE

Tel.: +33 4 92 94 42 00 Fax: +33 4 93 65 47 16

Internet

http://www.3gpp.org

Copyright Notification

No part may be reproduced except as authorized by written permission.The copyright and the foregoing restriction extend to reproduction in all media.

2008, 3GPP Organizational Partners (ARIB, ATIS, CCSA, ETSI, TTA, TTC).

All rights reserved.

Contents

5Foreword

1Scope62References63Definitions, symbols and abbreviations73.1Definitions73.2Symbols73.3Abbreviations74Considerations related to Dual-Cell HS-DSCH operation84.1Co-existence with legacy UEs84.2Carrier allocation84.2.1Anchor carrier and supplementary carrier84.2.2Cell definition84.2.3Sector definition84.2.4Time reference84.2.5Active set84.3Physical channel considerations94.3.1Allocation of common channels94.3.2Control channel structures94.3.2.1Uplink94.3.2.1.1HS-DPCCH94.3.2.2Downlink104.3.2.2.1HS-SCCH104.4Impact on system operation and procedures104.4.1L1/L2/L3 procedures104.4.1.1 Dynamic supplementary carrier enabling/disabling at the Node B104.4.1.2Mobility issues104.4.1.2.1Active set change, Serving cell change and Measurement reporting104.4.1.3 Fast power control114.4.1.4 CPC114.4.2UE capabilities114.5Scheduling considerations114.5.1 Joint vs. Disjoint Queues124.5.2Joint vs. Disjoint Scheduling124.5.3Joint vs. Disjoint HARQ retransmissions125Performance evaluation125.1Outline of performance evaluation methodology125.1.1Simulation procedure125.1.2Performance evaluation scenarios125.1.3Simulation assumptions125.1.3.1Bursty traffic145.1.3.2Full buffer traffic and balanced load between two carriers155.1.3.3Full buffer traffic and imbalanced load between two carriers155.1.4Evaluation metrics155.2Performance evaluation results165.2.1Simulation results and analysis provided by Source 1 [7]165.2.2Simulation results provided by Source 2 [8]215.2.3Simulation results provided by Source 3 [9]285.2.4Discussion on the difference in the simulations results305.2.5HS-DPCCH Cubic Metric Analysis305.2.5.1CM analysis305.2.6Alternate CM analysis335.2.6.1Maximum Cubic Metric with dual HS-DPCCH configurations355.2.6.2Impact on Coverage455.2.6.2.1 Cubic Metric Increase Due to DC-HSDPA455.2.6.2.2 Uplink Beta Gain Settings495.2.6.2.3 Link Budget Impact Due to DC-HSDPA495.2.6.3Simulation Results when E-DCH is not transmitted or configured496Impacts506.1Impact on implementation and complexity506.1.1UTRAN506.1.2UE506.1.2.1 DC-HSDPA High Level Requirements506.1.2.2 UE Receiver Types516.1.2.3 High Level UE Receiver Block Diagram526.1.2.4RF and Digital Front End536.1.2.5Base-band Detector566.1.2.5.1 LMMSE Processing576.1.2.6 Base-band Decoder586.1.2.7 UE Transmitter596.1.2.8 Conclusions on UE Complexity596.2Impact on specifications616.2.1Impact on L1 specifications616.2.2Impact on RRC specifications616.2.2.1URA_PCH and CELL_PCH states616.2.2.2Number of carriers from sectors in the active set616.2.2.3Channel assignment616.2.2.4Intra-frequency and Inter-frequency measurement616.2.2.4.1Reporting event 1D: Change of best cell616.2.2.4.2Event 2a: Change of best frequency616.2.3Impact on UE RF requirements616.2.3.1 New DL reference measurement channel626.2.3.2 UE maximum output power626.2.3.3 Reference Sensitivity Level626.2.3.4 Maximum Input Level636.2.3.5 Adjacent Channel Selectivity (ACS)636.2.3.6 In band blocking636.2.3.7 Narrowband blocking636.2.3.8 Out of band blocking646.2.3.9 Intermodulation Characteristics646.2.3.10 In-band ACS646.2.3.11 Spurious Emissions646.2.4Impact on UE demodulation performance requirements646.2.5Impact on NodeB RF requirements656.2.5.1 Frequency bands and channel arrangement656.2.5.2 Transmitter characteristics656.2.5.3 Receiver characteristics657Conclusion65Annex A: Proportional Fair Schedulers66Change history66

Foreword

This Technical Report has been produced by the 3rd Generation Partnership Project (3GPP).

The contents of the present document are subject to continuing work within the TSG and may change following formal TSG approval. Should the TSG modify the contents of the present document, it will be re-released by the TSG with an identifying change of release date and an increase in version number as follows:

Version x.y.z

where:

xthe first digit:

1presented to TSG for information;

2presented to TSG for approval;

3or greater indicates TSG approved document under change control.

ythe second digit is incremented for all changes of substance, i.e. technical enhancements, corrections, updates, etc.

zthe third digit is incremented when editorial only changes have been incorporated in the document.

1Scope

The present document is intended to capture findings produced in the context of the Feasibility Study on Dual-Cell HSDPA operation [1].

The work under this study item aims at assessing the feasibility, benefits and complexity of combining network radio resources (i.e. cells) to achieve enhanced user experience and enhanced user experience consistency. The assessment focuses on scenarios with the following constraints: The dual cell operation only applies to downlink HS-DSCH. The two cells belong to the same Node-B and are on different carriers. The two cells do not use MIMO to serve UEs configured for dual cell operation. Primary priority: The two cells operate on adjacent carrier frequencies in the same frequency band. Other allocations can be considered with lower priorityIn order to characterize benefits of Dual-Cell HSDPA operation, possible enhancements of performance throughout the cell and in particular in the outer area of the cell coverage are evaluated considering:

UE receiver impairments caused by the implementation of dual-cell operation,

Node B scheduler architecture (per carrier or joint scheduler),

Load balancing when coupled with joint scheduling vs. per carrier scheduling.

Furthermore, impacts on implementation and complexity within the UTRA and UE, impacts systems operation (e.g. UL controlchannel coverage and operation of legacy UE), and impacts on the core specifications due to introducing Dual-Cell HSDPA operation are identified. 2References

The following documents contain provisions which, through reference in this text, constitute provisions of the present document.

References are either specific (identified by date of publication, edition number, version number, etc.) or nonspecific.

For a specific reference, subsequent revisions do not apply.

For a non-specific reference, the latest version applies. In the case of a reference to a 3GPP document (including a GSM document), a non-specific reference implicitly refers to the latest version of that document in the same Release as the present document.

[1]3GPPTD RP-080228: "Feasibility Study on Dual-Cell HSDPA operation".

[2]3GPPTR21.905: "Vocabulary for 3GPP Specifications".[3]R1-081706 Simulation Assumptions for DC HSDPA Performance Evaluations[4] R1-081546, Initial multi-carrier HSPA performance evaluation, Ericsson, 3GPP TSG-RAN WG1 #52bis, April, 2008.

[5] R1-081361, System Benefits of Dual Carrier HSDPA, Qualcomm Europe, 3GPP TSG-RAN WG1 #52bis, April, 2008.

[6] R1-081706, Simulation Assumptions for DC HSDPA Performance Evaluations, Qualcomm Europe, Ericsson, Nokia, NSN, 3GPP TSG-RAN WG1 #53bis, May 2008.

[7] R1-082094, Text proposal for TR on simulation results (initial submission), Qualcomm Europe, 3GPP TSG-RAN WG1 #53bis, May 2008.

[8] R1-082135, System simulation results for DC-HSDPA operation, Ericsson, 3GPP TSG-RAN WG1 #53bis, May 2008.

[9] R1-081903, Initial simulation results for dual cell HSDPA operation, Nokia, 3GPP TSG-RAN WG1 #53bis, May 2008.

[10] Data Networks, Dimitri P. Bersekas and Gallager, 2nd edition, Prentice Hall, 1992.

[11] 3GPP TR 25.876 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Multiple Input Multiple Output in UTRA; (Release 7) V7.0.0 (2007-03)[12] 3GPP TS25.101 User Equipment (UE) radio transmission and reception (FDD).3Definitions, symbols and abbreviations

Delete from the above heading those words which are not applicable.

Subclause numbering depends on applicability and should be renumbered accordingly.

3.1Definitions

For the purposes of the present document, the terms and definitions given in [2] and the following apply. A term defined in the present document takes precedence over the definition of the same term, if any, in [2].Definition format

: .

example: text used to clarify abstract rules by applying them literally.

3.2Symbols

For the purposes of the present document, the following symbols apply:

Symbol format

3.3Abbreviations

For the purposes of the present document, the abbreviations given in [2] and the following apply. An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in [2].Abbreviation format

DC-HSDPADual-Cell HSDPA4Considerations related to Dual-Cell HS-DSCH operation 4.1Co-existence with legacy UEsLegacy UE operation will not be impacted by the introduction of the DC-HSDPA in the system. In particular, it should still be possible to operate a UE in MIMO mode or Tx diversity mode on either of the two carriers, while another UE could be in DC-HSDPA mode using these two carriers.4.2Carrier allocation

A UE in DC-HSDPA operation is able to simultaneously receive HSDPA traffic over two downlink carrier frequencies transmitted in the same frequency band from a single serving sector and to transmit on one uplink carrier frequency. The uplink carrier for a DC-HSDPA UE is not strictly tied to one of the two downlink carriers.4.2.1Anchor carrier and supplementary carrier

Anchor carrier: A UEs anchor carrier has all the physical channels including DPCH/F-DPCH, E-HICH, E-AGCH, and E-RGCH.Supplementary carrier: During dual carrier operation in CELL_DCH, the UEs supplementary carrier is the downlink carrier which is not the UEs anchor carrier.4.2.2Cell definition

[2] defines a Cell as a Radio network object that can be uniquely identified by a User Equipment from a (cell) identification that is broadcasted over a geographical area from one UTRAN Access Point. In DC-HSDPA, a cell means a radio network object representing a combination of a carrier and a geographical area.

4.2.3Sector definition

[2] defines a sector as a sub area of a cell. All sectors within one cell are served by the same base station. A radio link within a sector can be identified by a single logical identification belonging to that sector.

[2] implies that a sector refers to a geographical area of coverage. The sector nomenclature was introduced early in the WCDMA development. Since, it does not coexist with the way 3GPP specifications have evolved; this TR alters the sector definition to be associated with one or more cells on different carriers covering the same geographical area.

(Note: If the redefinition of a sector is not agreeable, we need a new term to carry this definition.)4.2.4Time reference

As stated in [1], the two cells in a multi cell sector belong to the same Node B.

As stated in [1], the two cells in a multi cell sector are transmitted using the same antenna(s).

The two carriers can have the same time reference and their downlinks can be time aligned in the sense that they share the same Tcell value. This simplifies the design and the downlink/uplink timing relationships. As a result there is only one (DPCH per UE.

4.2.5Active set

The active set definition might have to be updated for DC-HSDPA operation. Two possible options have been identified:

1. The active set is the aggregate of the legacy single carrier active set on the anchor carrier and the serving cell on the supplementary carrier.

2. The presence of a supplementary carrier can be disregarded in the active set definition, i.e. the active set size is not affected by the presence or absence of a supplementary carrier.4.3Physical channel considerations

There are no restrictions of channel operation on the anchor carrier. On the supplementary carrier, the UE can only monitor DL HSDPA related channels.

4.3.1Allocation of common channels

A carrier which is an anchor carrier for one or more UEs must transmit all the common control channels. It may be useful to allow a mode of operation where a carrier which is not the anchor carrier for any UE does not need to transmit common control channels except for the pilot and maybe the synchronization channels.4.3.2Control channel structures4.3.2.1Uplink

4.3.2.1.1HS-DPCCHA few design options exist to modify HS-DPCCH for the purpose of carrying ACK/NACK and CQI for both carriers. Other options might be considered as well.A second HS-DPCCH channel similar to the existing one can be transmitted in parallel to the usual HS-DPCCH. It is assumed that a maximum of 1 DPDCH is supported on the uplink.

In the case when no DPDCH is configured in uplink, the two HS-DPCCHs can be I/Q multiplexed on the same channelization code that is used today for HS-DPCCH.

In the case when one DPDCH is configured in uplink, the two HS-DPCCHs can also be I/Q multiplexed on a single channelization code if a channelization code is chosen where both branches are available, i.e. a different channelization code than the one that is used today for HS-DPCCH. As an alternative, the two HS-DPCCHs can be allocated to two different channelization codes, either on the same branch or on different branches.

These design options can be summarized as follows:

Option A:

N_max_dpdch = 0, where N_max_dpdch is the maximum number of configured uplink DPDCHs The UE sends the 1st HS-DPCCH and 2nd HS-DPCCH on the same channelization code.

The channelization code is the same code as that is currently used for the legacy HS-DPCCH.

The 1st HS-DPCCH is sent on the Q branch and the 2nd HS-DPCCH is sent on the I branch.

Option B:

N_max_dpdch = 1

The channelization code for the 1st HS-DPCCH is the same code as that is currently used for the legacy HS-DPCCH.

The channelization code for the 2nd HS-DPCCH is a different code from that is currently used for the legacy HS-DPCCH.

The 1st HS-DPCCH is sent on the Q branch and the 2nd HS-DPCCH is sent on the I or Q branch.

When the UE is configured back to SC-HSDPA mode, it switches to the channelization code/branch that is currently used for the legacy HS-DPCCH.

Option C:

N_max_dpdch = 1

The UE sends the 1st HS-DPCCH and 2nd HS-DPCCH on the same channelization code.

The channelization code is not the same code as that is currently used for the legacy HS-DPCCH.

The 1st HS-DPCCH is sent on the Q branch and the 2nd HS-DPCCH is sent on the I branch.

When the UE is configured back to SC-HSDPA mode, it switches to the channelization code/branch that is currently used for the legacy HS-DPCCH.

Option D:

N_max_dpdch = 1

The UE sends the 1st HS-DPCCH and 2nd HS-DPCCH on the same channelization code.

The channelization code is not the same code as that is currently used for the legacy HS-DPCCH.

The 1st HS-DPCCH is sent on the Q branch and the 2nd HS-DPCCH is sent on the I branch.

When the UE is configured back to SC-HSDPA mode, it continues to use the new channelization code/branch that was assigned to 1st HS-DPCCH.4.3.2.2Downlink

4.3.2.2.1HS-SCCH

Both the anchor and supplementary carriers can have disjoint HS-SCCH channels. In this case the coding of HS-SCCH does not need to be changed. In order to not restrict the scheduler, the UE should preferably monitor up to 4 HS-SCCH codes on each carrier, as in the single carrier case, assuming HS-SCCH is transmitted in both carriers.4.4Impact on system operation and procedures

4.4.1L1/L2/L3 procedures

4.4.1.1 Dynamic supplementary carrier enabling/disabling at the Node B

From e.g. the UE battery point of view, it might be beneficial for the Node B to be able to enable and disable the supplementary carrier based on the downlink traffic and channel conditions.4.4.1.2Mobility issues4.4.1.2.1Active set change, Serving cell change and Measurement reporting

Although the decision will still reside in the network, there are many possible ways the UE can assist the management of the active set and the serving cell:

Option 1:

The UE monitors and reports events based on the anchor carrier only. This is the simplest scheme as it takes the existing mechanism and ignores the supplementary carrier.

Option 2:

The UE monitors both carriers and reports when the events are triggered on the anchor carrier. This is an enhancement to the current scheme where the UE reports the measurement from both carriers when the triggers are triggered on the anchor carrier. Even though this is an enhancement, it still does not catch all the possible trigger points as the triggers are not based on the supplementary carrier.

Option 3:

The UE monitors both carriers and reports when the events are triggered on either carrier. This mechanism allows the network to receive all the information. The problem is that it can go too far as it could be triggering double the numbers of events needlessly. The reported measurements could be for the anchor carrier only or for both carriers, whenever any of the events is triggered.

Option 4:

The UE monitors both carriers and reports throttled events from both carriers. This proposal tries to get most of the gains without burdening the network with superfluous reports. The simplest way of achieving this is to throttle the events from both carriers, in order to avoid sending duplicate messages for similar triggers happening within a short time frame from each other. The event can still be triggered by one carrier changing conditions, however when the other carrier changes as well, it would probably change in a short amount of time that would be caught by the throttling mechanism.

Option 5:

The UE monitors both carriers and reports events based on the combination of both carriers. This option can be effective, efficient and flexible. It maximizes the inherent value of reports (triggers) rather than only measurements (contents) so that UE can report from a performance standpoint and UTRAN can decide on handover-off or serving cell change from a resource standpoint (i.e., do not dilute the inherent value of reports by over-reporting). It can consider the aggregate merit of cells across carriers for HS-DSCH serving cell changes rather than the individual carrier merit (e.g. total effective throughput achievable for expected CQIs per carrier). It also can minimize reporting (signalling) overhead but do not under-report. Finally, it removes relieves the network from guessing when to perform mobility procedures.4.4.1.3 Fast power control

Uplink power control can operate in such a way that the UE uplink transmit power is controlled by the network through an F-DPCH transmitted on the anchor carrier. Similarly, downlink power control can operate such that the power of the F-DPCH on the anchor carrier is controlled by the UE sending TPC commands on the UL DPCCH.

4.4.1.4 CPC

HS-SCCH-less operation can be restricted to the anchor carrier, while UE DTX/DRX can be carried out taking both carriers into account (details are FFS).4.4.2UE capabilities

Support for DC-HSDPA operation would be a UE capability.4.5Scheduling considerationsThe serving cells on both carriers belong to the same sector.

4.5.1 Joint vs. Disjoint Queues

The downlink queues at the Node B could be operated in a joint or disjoint manner for the two carriers. The simulations assumption of [3] is that the queues are joint.

4.5.2Joint vs. Disjoint Scheduling

Whether the scheduling over the two DL carriers is joint or disjoint, does not impact the specifications. The simulations assumption of [3] is that the scheduling is joint and using a proportional fair algorithm .4.5.3Joint vs. Disjoint HARQ retransmissionsThe simulations assumption of [3] is that HARQ retransmissions are assumed to go on the same carrier as the first transmission.

5Performance evaluation5.1Outline of performance evaluation methodology5.1.1Simulation procedure5.1.2Performance evaluation scenarios

5.1.3Simulation assumptions

In general, the parameters listed below are the same as those in TR 25.848 and TR 25.896.

Some parameters or algorithms will be left open for each company to pick its favourite. These are marked with an asterisk (*).ParametersValues and comments

Cell LayoutHexagonal grid, 19 Node B, 3 sectors per Node B with wrap-around

Inter-site distance1000 m

Carrier Frequency2000 MHz

Path LossL=128.1 + 37.6log10(R), R in kilometers

Log Normal Fading Standard Deviation: 8dB

Inter-Node B Correlation: 0.5

Intra-Node B Correlation:1.0Correlation Distance: 50m

Max BS Antenna Gain14 dBi

Antenna pattern

= 70 degrees, Am = 20dB

Channel Model PA3, VA3Fading across carriers is independent for non adjacent carriers.

(*) Two fading models for adjacent carriers:

- Fading across carriers is completely uncorrelated.

- Fading correlation across carriers is modeled using some practical approach (optional)

- Fading across carriers is completely correlated

Penetration loss10 dB

CPICH Ec/Ior-10 dB

HS-DSCH Up to 15 SF 16 codes per carrier for HS-PDSCH

(*) Power allocation:

- Total available power for HS-PDSCH and HS-SCCH is 70% of Node B Tx power, with HS-SCCH transmit power being driven by 1% HS-SCCH BLER, or

- Total available power for HS-PDSCH is 75% of Node B Tx power, with a fixed HS-SCCH transmit power and an ideal decoding, or

(*) HS-PDSCH HARQ: Both chase combining and IR based can be used. Maximum of 4 transmissions with 10% target BLER after the first transmission. Retransmissions are of highest priority.

HS-DPCCH 9 slot CQI delay

CQI bias is 0 and CQI estimation noise is Gaussian with 1 dB std

(*) CQI quantization may or may not be modeled

Error-free CQI and ACK decoding

UE Antenna Gain0 dBi

UE noise figure9 dB

Thermal noise density-174 dBm/Hz

UE capabilities15 SF 16 codes capable per carrier

UE Receiver TypeType 2 and Type 3 for both single carrier and DC-HSDPA (*) Realistic C/I estimation

Maximum Sector

Transmit Power43 dBm per carrier

Other Sector Transmit Power(*) If OCNS=1, all other sectors always transmit at full power;

(*) If OCNS=0, other sectors transmit at full power only when they have data.

TimingThe two carriers have the same time reference and their downlinks are synchronized.

Serving cellThe serving cells on both carriers belong to the same sector.

Traffic modelFull buffer and Bursty Traffic Model (as specified in Section 5.1.2)

Queuing and SchedulingJoint-queue (**) and Proportional Fair (e.g. as specified in Annex A)

Traffic distribution Uniform over the area

Number of UEs per sector1, 2, 4, 8, 16, 32, 64

In addition, other number of UEs per sector can also be considered.

(*) Parameters or algorithms possibly different between companies.

(**) The data on both carriers in DC HSDPA share the same queue at the Node B.5.1.2Traffic Models

There are two types of traffic: full buffer and bursty traffic.

Full buffer traffic assumes that each user always has data.

The following simple model is used for bursty traffic: the burst size is log-normally distributed as in FTP traffic model described in TR 25.876 but with the parameters described in the following table. There is no underlying transport protocol modeled. The inter-burst time is the time between the arrival of two consecutive bursts.

ComponentDistribution

Parameters

PDF

File size (S)Truncated LognormalMean = 0.5 Mbytes

Std. Dev. = 0.1805 Mbytes

Maximum = 1.25 Mbytes

Inter-burst time ExponentialMean = 5 sec, 20 sec

5.1.3Simulation scenarios and performance metrics5.1.3.1Bursty traffic

Assuming there are two carriers and altogether 2*N users per sector. In the single carrier system, there are N users in each carrier. In DC HSDPA, all 2*N users use dual carriers.

The following performance metrics should be compared between the single-carrier system and DC HSDPA:

Average burst rates at different number of users (N)

The burst rate is defined as the ratio between the data burst size in bits and the total time the burst spent in the system

The total time the burst spent in the system is the time difference measured between the instant the data burst arrives at the Node B and the instant when the transfer of the burst over the air interface is completed.

The total time the burst spent in the system is equal to the sum of the transmission time over the air and the queuing delay.

Total system throughput

Normalized and un-normalized user throughput distribution (CDF)

5.1.3.2Full buffer traffic and balanced load between two carriers

Assuming there are two carriers and altogether 2*N users per sector. In the single carrier system, there are N users in each carrier. In DC HSDPA, all 2*N users use dual carriers.

The following performance metrics should be compared between the single-carrier system and DC HSDPA:

Sector throughput at different number of users (N)

Normalized and un-normalized user data rate distribution (CDF)

User data rate gain at different user data rate percentiles: This would be the user throughput improvements as a function of the user-quantile (relative improvement of average per-user throughput over user-quantile, e.g. by how much did the throughput of the worst 10% of users improve). This is metric can demonstrate any cell edge user performance enhancement

Average user throughput as a function of average sector throughput.

5.1.3.3Full buffer traffic and imbalanced load between two carriers

This is an optional scenario.

Without multicarrier operation, moving users across carriers is a slow procedure. Even if the network equalizes the number of users across carriers, in real life, there is no sustained full buffer traffic. The traffic for a particular user is bursty and the number of users simultaneously receiving packets in each carrier at any given time can be different. The gains in these situations can be shown by studying full buffer traffic with imbalanced number of users across carriers.

Assuming there are two carriers and altogether 2*N users per sector, let M be the number of users in the first carrier and K the number of users in the second carrier, where M+K=2*N and M(K. In DC HSDPA, all 2*N users use dual carriers.

The following performance metrics should be compared between the single-carrier system and DC HSDPA:

Sector throughput at different total number of users (2*N) and at different user-carrier association (M,K) with the same total number of users,

Normalized and un-normalized user data rate distribution (CDF)

User data rate gain at different user data rate percentiles

Average user throughput as a function of average sector throughput

5.1.4Evaluation metrics5.2 Performance evaluation resultsThe following example explains this terminology. Consider the case when we have 8 users per sector. By this, we mean that there are 8 users in 10 MHz. When we consider balanced load between carriers, we will compare performance when 4 of the users are on each cell (5 MHz) with the performance when all 8 users are capable of receiving data on both cells (10 MHz). We refer to the former as 2x-single cell (2x-SC HSDPA) case and the latter as the dual cell (DC-HSDPA) case. Note that in the 2x-SC HSDPA case, the load is balanced across carriers.

5.2.1Simulation results and analysis provided by Source 1 [7]5.2.1.1Choice of parameter values

In this subsection, the choice of optional parameter values in Section 5.1.3 is listed in the table below.

ParametersComments

Channel ModelPA3

UE Receiver TypeType 3 (LMMSE with RxD)

HS-DSCH PowerMaximum Power = 70% of Node B transmit power

HS-SCCH power decided by a 1% HS-SCCH BLER

HS-DSCH power margin driven by an outer loop (10% BLER after 1st Tx, Max 4 HARQ Transmissions)

Other Sector Transmit PowerOCNS = 1 (all other sectors always transmit at full power)

Fading Across CarriersUncorrelated

Channel EstimationRealistic

5.2.1.2Gains with full buffer traffic under balanced load

Figure 1 shows the improvement of average user throughput due to dual cell HSDPA as a function of sector throughput. For both 2xSC and DC-HSDPA, we compare the average user throughputs at the same number of users per sector. As we can see, the dual cell gain is more pronounced at low loads. This is because multi-user diversity gain is larger in DC-HSDPA as there are more users to choose from at each scheduling instance. As the load increases, the gains from multi-user diversity and joint scheduling decrease. At 2 users per sector, the gain is around 25%. At 32 users per sector, it is around 7%.

Figure 1 Average user throughput as a function of sector throughput. Figure 2 shows the CDF of user throughputs for 16 users per sector. We see that the percentage gain for low geometry users is higher than that for high geometry users. Figure 3 shows the CDF of normalized user throughput (fairness curves). We can see that DC HSDPA is fairer than 2xSC-HSDPA.

Figure 2, 3: User throughput CDF and fairness curve (16 users per sector)

This behaviour can be seen more clearly when we plot user throughput gains as a function of user percentile. Essentially, from the CDF of user throughput, we identify the 10%, 20%, , 90%-ile throughputs from both the 2x-SC and DC-HSDPA curves and compare them.Figure 4 shows us the user throughput gains as a function of user percentile. At low percentiles (analogous to low geometries), the gains are higher than at high percentiles (high geometries). This is because low geometry users see a higher variation in their proportional fair metric (see Appendix of Error! Reference source not found.). Higher geometry users will see a lower variation of this metric, given that they are in more likely to be in the non-linear region of the Shannon curve.

Figure 4. User throughput gains of DC-HSDPA over 2xSC-HSDPA as a function of user percentile

Figure 5 shows the gain in sector throughput as a function of number of users per sector. Again, as we can see, DC-HSDPA gain is more pronounced at low loads. At 2 users per sector, the gain in sector throughput is 25%. At 32 users per sector, it is 7%.

Figure 5 Capacity gain from DC HSDPA over 2xSC-HSDPA5.2.1.3DC HSDPA gains with bursty traffic

As seen above, compared with two single carriers each with N users, DC HSDPA with 2*N users results in a small gain in terms of sector capacity with full buffer traffic data. However, with bursty traffic, DC HSDPA provides a significant gain in terms of latency reduction. A more intuitive performance metric is the burst rate [6] defined as the ratio between burst size and the time taken to transfer the burst over the air interface from the time it arrives at Node B. The gain can be seen from queuing analysis and system simulations.

5.2.1.4

Queuing analysis of DC HSDPA latency reduction and burst rate increase

The following analysis was presented in [5].

Lets assume a M/G/1 queuing system. The service rate can be random with any distribution. The arrival process is assumed to be memoryless[6], namely, the inter-arrival times are exponentially distributed. This model captures many features in the bursty traffic services in the HSDPA systems.

For one single carrier, lets denote the arrival rate is and departure rate is When we have two carriers and twice the number of users, namely, i.e., the same number of users per cell (per sector per carrier), we have another M/G/1 system with arrival rate 2 and service rate 2. It is obvious that the actual service time of each burst is reduced by half in the alternative system. Therefore, to quantify the gain in the burst rate, we need to find the waiting time, which in turn depends on the queue length. If we compress the time resolution to half in the new M/G/1 system with 2 and 2, the queue length dynamic is exactly the same as in the original M/G/1 system with and . Therefore, the average queue length remains the same but the average waiting time is cut in half.

The same conclusion can be seen from the Kleinrock-Khinchin formula for M/G/1 queue[10]. The total time for a data burst in the system is

,

where is the second moment of the service time. As we can see, when both and doubled, is reduced to a quarter of its value and the total time in system is reduced by half. 5.2.1.5

Simulation results with bursty traffic

In [5], we had provided analysis and simulation results for burst rates for 2x-SC and DC HSDPA assuming a fixed burst size. In this document, we provide burst rate curves for the traffic model where the burst size follows a truncated log-normal distribution. Figure 6 shows the distribution of burst sizes.

Figure 6 Burst Size CDF

Figure 7 shows the CDF of burst rates for the 8 users per sector. Note that there are 8 users in 10 MHz for both 2xSC and DC-HSDPA. We see that there is a ~2x improvement in the burst rates with dual cell HSDPA compared to 2xSC-HSDPA.

Figure 7 Burst Rate CDF for 8 users per sector. The blue curve refers to the case when 4 of the users are on each cell (2xSC-HSDPA) while the red curve refers to the case when all 8 users are dual cell capable (DC HSDPA).

Figure 8 shows the number of users that can be supported as a function of the average burst rate per user. As the load increases, we see that the gains from DC HSDPA start to fall, as the queue length begins to increase and begins to resemble full-buffer. Note that the number of users per sector is proportional to the load seen by the scheduler. Please note that other cell powers are set to maximum, so partial loading effects are not seen in Figure 8. If partial loading is explicitly simulated, the burst rates will be much higher.

Figure 8 Burst rate performance with OCNS=1.

Figure 9 compares the sector throughput at the application layer for DC HSDPA with 2xSC-HSDPA. The application layer throughput is smaller than the physical layer throughput. Since we do not model TCP, the difference between the physical and application layer throughputs is only the header overhead. Relative Comparison between 2xSC-HSDPA and DC-HSDPA is independent of the overhead.

Read in conjunction with Figure 8, we see that while the burst rates have doubled, the sector throughput is the same for both. In other words, the burst is served faster in DC-HSDPA and therefore, there is more idle time in DC-HSDPA than in the 2x-SC HSDPA. As the number of users per sector increases beyond 64, the sector throughput curves will saturate.

Figure 9 Sector throughput as a function of users per sector

In summary, the simulations show:

For a given burst rate, DC HSDPA can support more than twice the number of users compared to 2x-single cell HSDPA at low loads. For instance, at a burst rate of 3.5 Mbps, the number of users supportable with DC HSDPA is more than twice the number that can be supported through 2xSC HSDPA.

At low to medium loads, for a given number of users, DC HSDPA can provide a doubling of the burst rate compared to 2xSC-HSDPA.5.2.2Simulation results provided by Source 2 [8]5.2.2.1Choice of parameter values

In this subsection, the choice of optional parameter values in Section 5.1.3 is listed in the table below.

ParametersComments

Channel ModelPA3

UE Receiver TypeType 2 (LMMSE without RxD), Type 3 (LMMSE with RxD)

HS-DSCH PowerMaximum Power = 70% of Node B transmit power

HS-SCCH power decided by a 1% HS-SCCH BLER

HS-DSCH power margin driven by an outer loop (10% BLER after 1st Tx, Max 4 HARQ Transmissions)

Other Sector Transmit PowerOCNS = 0 (multicell simulation with active users in each cell where the interference level is a consequence of the current situation in the other sectors)

Fading Across CarriersCorrelated

Channel EstimationCQI estimation error of 1 dB

5.2.2.2Simulation results for Bursty traffic

If a user is downloading traffic burst 1 and burst 2 arrives before burst 1 is finished there are several ways of dealing with this situation. In this investigation we start the download of burst 2 as soon as it arrives. Burst 1 and 2 will share the available resource until burst 1 (or 2) is finished.As a consequence of the traffic model there is a straightforward mapping between the number of users in a sector and the offered load in bits/s/sector. Each user contributes with 200kbit/s to the offered load. We use the offered load on the axes instead of number of users since it makes the result a bit more general. Note that in some cases results will depend on the simulation time, e.g. for an unstable system. In these results a 57 sector system was simulated for 5 minutes.

The recommended load of 64 users per cell can not be handled by the system in any of the investigated scenarios. Where a really high load was interesting a load of 50 users (10 Mbit/s/sector) was used instead.The results are shown in Figure 10 through Figure 13.

Figure 10 shows that for all load levels, DC-HSDPA gives roughly twice as high average user throughput as two single carriers with the corresponding receiver structure. This is a consequence of the better low load properties of DC-HSDPA compared to two single carriers. It is much more unlikely that there is a build-up of files in a sector when DC-HSDPA is used, which leads to higher performance also for the higher loads.

When we study the 10 and 90 percentiles in Figure 11 and Figure 12 we realize that this performance increase is valid for all users in the system.

A system can be said to be stable when the output of the system is equal to the input. If we plot the transmitted bits and the bits that arrived to the system as a function of the average number of users as in Figure 13, we get a clue whether a certain scenario results in stable operation. From this graph we can guess that a load of 32 users per sector (6.4 Mbit/s in offered load) is too much for systems without Rx diversity to handle.

Figure 10: User throughput vs sector throughput for Bursty traffic

Figure 11: 10 percentile user throughput vs sector throughput for Bursty traffic

Figure 12: 90 percentile user throughput vs sector throughput for Bursty traffic

Figure 13: Transmitted bit vs offered bits for Bursty traffic

5.2.2.3Simulation results for Full buffer traffic and balanced load between two carriers

If there is an even number of users in a sector, 2*n where n = 0, 1, 2, 3 ., each carrier will have exactly n users. In a situation where the number of users in the sector is odd, 2*n +1 where n = 0, 1, 2, 3 ., one randomly selected carrier will have n+1 users and the other one will have n users.

The following average numbers of users have been simulated: 0.25, 0.5, 1, 2, 4, 8, 16, 32.

The results are shown in Figure 14 through Figure 19.

We see that there is large difference in both average user and sector throughput depending on the receiver type. Receiver type 3 gives ~30% higher system capacity than type 2. At low number of users the DC-HSDPA solution clearly outperforms the corresponding double single carrier solution.

In Figure 17, Figure 18 and Figure 19, the 10/50/90 percentile user throughput for the different scenarios is normalized with the 10/50/90 percentile user throughput of double single carriers with receiver type 2.

Figure 14: User throughput vs sector throughput for Full buffer traffic

Figure 15: Average user throughput vs average number of users per sector for Full buffer traffic

Figure 16: Average sector throughput vs average number of users per sector for Full buffer traffic

Figure 17: 10 percentile throughput gain vs average number of users per sector for Full buffer traffic

Figure 18: 50 percentile throughput gain vs average number of users per sector for Full buffer traffic

Figure 19: 90 percentile throughput gain vs average number of users per sector for Full buffer traffic

5.2.3Simulation results provided by Source 3 [9]System simulation results in Pedestrian A channel are shown in Figure 20 and Figure 21 assuming fairness factors 0.1 and 0.001, respectively. Actually first fairness factor is such that scheduling is pretty close to round robin. Results are shown both with and without receiver diversity. As can be seen in the results in Figure 21 receiver diversity reduces gain due to multicarrier somewhat. In this case also sector throughput seems to be higher for 2xSC with receiver diversity than for multicarrier withour receiver diversity except the case where there is only one user per sector. Multicarrier gain reduces as number of users per sector increases. Highest gain of roughly 100% is achieved in a special case when there is only one user per sector since in that case multicarrier user can utilize both carriers all the time.

Parameters used in simulation are presented in detail in Annex A of [9] and are compliant with the scenario agreement in [6]. Full buffer traffic is used in all simulations.

Figure 20 Average sector throughput in case of fairness factor 0.1

Figure 21 Average sector throughput in case of fairness factor 0.0015.2.3.1Annex A in [9]System simulation parametersParameter Value

Cellular system WCDMA HSDPA

Carrier bandwidth 5 MHz

Number of carriers 2

Carrier 1 frequency 2150 MHz

Carrier 2 frequency 2155 MHz

Sectors per cell 3

Site-to-site distance 1000 m

Minimum BS and MS separation 35 m

HS-PDSCH transmit power 75 %

CPICH transmit power 10 %

Thermal noise -99 dBm

BS total transmit power 43 dBm

Propagation model 16 + 37.6 log10(d[m])

Correlation between sites 0.5

Correlation between sectors 1.0

Standard deviation of slow fading 8 dB

Mobile speed 3 km/h

Mobile receiver type LMMSE chip equalizer

ITU channels Extended Ped A

Number of multicodes 15 (variable)

CQI set 0.5 QPSK, 0.75 QPSK, 0.5 16QAM, 0.75 16QAM

AMC feedback delay 3 TTIs

AMC packet-error-rate target 50 %

Fast HARQ scheme Chase combining

HARQ processes 6

HARQ transmissions 4

Packet scheduler proportional fair

Traffic model Full buffer

5.2.4Discussion on the difference in the simulations results

In the full buffer results, the user and sector throughput provided by different sources are fairly close in comparable cases. The reasons for the minor difference include the following: different models on receiver performance; difference in the scheduler including the fairness criteria, channel sensitivity, assumptions on the channel fading correlations between the carriers and whether multiple users can be scheduled for the same TTI. For example, all the Sources use Proportional Fair (PF) scheduler. But an extra parameter of fairness factor is used by Source 3. Longer time constant in the PF-Scheduler is used by Source 1 (2250 slots, or 1.5 seconds) than Source 2 (192 slots) and therefore higher multi-user diversity seen in results by Source 1.

In the results with bursty traffic, the burst rates reported by Source 2 are higher than those by Source 1 for the comparable cases although the results converge with a large number of users. The main reason, apart from listed above, is the interference from the non-central sectors. In the simulations provided by Source 1, N users (N=1,2,4,8,16,32,64) are dropped uniformly to the central sector. All the other sectors are assuming to transmit with full power all the time according to the assumption of OCNS=1 [6]. The reported results are the average performance over multiple drops. In the simulation provided by Source 2, 57*N users are dropped uniformly to the entire 57-sector system. The transmit power in the non-central sector is explicitly simulated. Therefore, with small to medium number of users per sector, the data rates seen by Source 2 will be higher since non-central sectors are not always transmitting with full power. When the number of users becomes large, the difference between the two simulations shrinks. Considering all the simulation results provided by various sources, the following common trends can be observed:

For full buffer traffic:

DC HSDPA results in user throughput and sector throughput gains. Such gains are more significant with small number of users per sector and decrease with number of users.

Low geometry users gain more in terms of throughput than high geometry users.

For bursty traffic:

DC HSDPA results in a doubling of burst rates with low to medium loads, even after normalizing the number of users per 5 MHz.

At low to medium loads, for a given burst rate, DC HSDPA can support more than twice the number of users when compared to 2xSC-HSDPA.

Such gain decreases when the load is so high that the queues of users in bad geometry start to build up.5.2.5HS-DPCCH Cubic Metric Analysis

In the following, we present a cubic metric impact analysis, due to transmission of the 2nd HS-DPCCH as described in 4.3.2.1.1.1. The analysis assumes that a maximum of 1 dedicated channel is supported on the uplink. 5.2.5.1CM analysisAs an initial analysis, we generate CM values to investigate the impact of introducing new HS-DPCCH (HS-DPCCH2).

Table 1 summarizes the channelization code and the gain factor values of the reference channel configuration which are used in the simulation. Table 2 describes the channelization code and gain factor values of HS-DPCCH2.

Table 1: Channel configuration of the reference channelsChannelChannelization codeGain factor

Nmax-dpdch=0DPCCH(Q,256,0)15

E-DPCCH(I,256,1)24

E-DPDCHSF4=(I,4,1)

SF2x2=((I,2,1) (Q,2,1)

SF2x2+SF4x2 = ((I,4,1),(Q,4,1),(I,2,1),(Q,2,1)){17,27,47,67,84}

HS-DPCCH(Q,256,33){5,12,24,38}

Nmax-dpdch=1DPCCH(Q, 256,0)15

E-DPCCH(I, 256,1)24

DPDCH(I, 64, 16)21

E-DPDCHSF4=(I,4,2)

SF2x2=((I,2,1) (Q,2,1){17,27,47,67,84}

HS-DPCCH(Q, 256,64){5,12,24,38}

Table 2: Channel configuration of the additional HS-DPCCHChannelChannelization codeGain factor

Nmax-dpdch=0HS-DPCCH2 mapped on I branch(I, 256,33){5,12,24,38}

HS-DPCCH2 mapped on Q branch(Q, 256,32){5,12,24,38}

Nmax-dpdch=1HS-DPCCH2 mapped on I branch(I, 256,63){5,12,24,38}

HS-DPCCH2 mapped on Q branch(Q, 256,63){5,12,24,38}

Figure 1, 2 and 3 shows the CM values in case of Nmax-dpdch=0 and figure 4 and figure 5 shows the CM values in case Nmax-dpdch=1. The blue line and the red line indicate CM values in each case and the green and the purple line indicate the CM increase compared to the case of no HS-DPCCH2.a) Nmax-dpdch=0

Figure 22: CM values in case of SF4

Figure 23: CM values in case of SF2x2

Figure 24: CM values in case of SF2x2+SF4x2a) Nmax-dpdch=1

Figure 25: CM values in case of SF4

Figure 26: CM values in case of SF4Observations

In the case of Nmax-dpdch=0, HS-DPCCH2 mapped to I branch results in the smaller CM increase compared to HS-DPCCH2 mapped to Q branch. In the case of Nmax-dpdch=1, HS-DPCCH2 mapped to Q branch results in the smaller CM increase compared to HS-DPCCH2 mapped to I branch. Assuming HS-DPCCH2 mapped to I branch for Nmax-dpdch=0 and HS-DPCCH2 mapped to Q branch for Nmax-dpdch=1,

The maximum CM value is 3.38 dB in the case of SF4, beta_ed 27 and beta_hs=5.

The CM increase is higher when one E-DPDCH is used. In some case, 1dB CM increase is observed.5.2.6Alternate CM analysis

The possible channelization code indices that can be used for this 2nd HS-DPCCH under the worst case scenarios (different N_max_dpdch) is shown in Table 3.

In the cubic metric analysis performed here, we have run Monte-Carlo simulations to compare cubic metrics of single HS-DPCCH and dual HS-DPCCH with different code and channel allocation for the second HS-DPCCH.

Depending on N_max_dpdch, the 1st HS-DPCCH is still transmitted as before on the following channelization codes:

N_max_dpdch = 0

Cch,256,33 on Q N_max_dpdch = 1

Cch,256,64 on Q

Irrespective of N_max_dpdch = 0 or 1, we transmit the 2nd HS-DPCCH on the following channelization codes: Cch,256,0 on I

Cch,256,1 on Q Cch,256,32 on I

Cch,256,33 on ITables 4, 5 and 6 list the different simulation parameter settings performed in this analysis. The results obtained are categorized into 35 cases for each TTI setting as shown in Tables 7, 8, 9 and 10. Table 11 provides a summary of the results obtained from all the simulations.Table 3: Worst Case Code consumption for different N_max_dpdchN_max_dpdchUL ChannelsCode UsageIQ

04 E-DPDCH (2SF2+2SF4) +

1 E-DPCCH +

1 DPCCH +

1 HS-DPCCHUsedE-DPDCH1 Cch,2,1

E-DPDCH3 Cch,4,1

E-DPCCH Cch,256,1DPCCH Cch,256,0

E-DPDCH2 Cch,2,1

E-DPDCH4 Cch,4,1

HS-DPCCH Cch,256,33

Avail. for HS2, Cch,256,n0n 63, n 11n 63, n 33

11 DPDCH +

2 E-DPDCH (2xSF2) +

1 E-DPCCH +

1 DPCCH +

1 HS-DPCCHUsedDPDCH Cch,4,1

E-DPDCH2 Cch,2,1

E-DPCCH Cch,256,1DPCCH Cch,256,0

E-DPDCH2 Cch,2,1

HS-DPCCH Cch,256,64

Avail. for HS2, Cch,256,n0n 63, n 11